(342ax) Epidemic Models with ‘Time Since Infection’: Efficient and Flexible Modeling Tools

Epidemic models are useful tools in the fight against infectious diseases, but their usefulness is limited (in part) by their ability to accurately describe the underlying disease dynamics. The most accurate epidemic models are typically the most computationally expensive, and “compartment models” offer the most popular compromise between speed and accuracy. In a compartment model, an infected person progresses through a series of artifical ‘stages’ or ‘compartments’ (e.g. exposed, infectious, recovered), and the resultant equations are easily solved by standard tools for ordinary differential equations. A more realistic model would describe disease dynamics as a continuous function of time since infection (TSI), but the computational cost of a TSI model is typically assumed to be large by comparison.

In this talk, we share our recent work on TSI models. First, we improve upon existing TSI models by using a ‘filter’ to partition the infection population into discrete compartments, as and when such measurements are necessary for informing policy decisions (e.g. predicting hospitalizations, deaths, etc.). Second, we provide a more efficient numerical method for solving the equations of a TSI model with spectral accuracy. Given this numerical approach, we find that TSI models are now cost-competitive with the standard ‘compartment’ strategy for many applications.